Monday, May 27, 2024

NVIDIA’s OVX and the Misconception of Digital Twins

Datamation content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.

We saw our first glimpse of what a true digital twin could be at NVIDIA’s GTC last month. For many of us, it cleared up a misconception we did not know we had and highlighted a new service, called NVIDIA OVX, which is also part of NVIDIA’s Omniverse cloud. 

NVIDIA OVX is a new server designed to facilitate the creation and maintenance of digital twins. The misconception is that once created, a digital twin is done — when in fact, creating the digital twin is only the smallest part of the effort. Assuring it remains a twin is critical to its anticipated use and viability for simulation.

Let me break it down for you:

Digital twins

A digital twin is different from a twin in the biological world. In biological terms, identical twins start to diverge as they develop and are born. They are different people. You could not ask one a question and expect them to reliably reflect what the other would say.

But digital twins must remain twins throughout their useful lives. Otherwise, they instead cease to be digital twins and instead become a representation of the past. Granted, historic representations also have a use, but they cannot be used as digital twins anymore or ever probably.

You do not have to have digital twins in the metaverse, either. For example, a metaverse representation of a video game might have no connection to reality at all or use a past representation of what was a digital twin to make for more realistic game play. 

But the metaverse that will use digital twins is designed to simulate the real world.

Why digital twins need to remain connected

If you are running a simulation, you need the virtual world to be as close to the real world as possible. For instance, if you are testing an autonomous car and the representation of the world you are testing in has no relation to the world it will operate in, you’ll likely introduce errors you won’t catch until the autonomous car is on the road. That process could be deadly.

Looking at robots, particularly large-scale autonomous robots, a simulation based on an old floor plan will result in, at best, confusion and breakdowns when walls, stairs, and other major elements are not where the robot expects them to be. Something as simple as an electrical wire or conduit on the floor that the robot’s sensors do not see could trouble the robot and result in an injury. In addition, given that when a robot has a problem, you’ll want to use the mirror image universe to both analyze the problem and correct it. If that universe is materially different from the real world, your conclusions and fix may make the problem worse not better. Because you are dealing with what is effectively a corrupted digital twin — not because the twin was changed, but because the real world is no longer represented by it. 

In fact, with digital twins, the twin can become corrupted when the real world is changed, and the digital twin is not. You do not have to corrupt the twin to break the model, you just must break its connection to reality.

See more: NVIDIA’s Isaac Sim: Preparing for the Coming Age Of Robotics


The OVX servers and their related Omniverse cloud instances are designed to not only help you create an initial digital twin, but to help link it to sensors in the real world that will keep them connected. This means they must be able to handle massive amounts of different kinds of input, apply that data to the rendering of the digital twin, and do all that in real-time. So the twin never ceases to be a twin and the resulting simulations can accurately predict future events and be used effectively to train autonomous machines.  

This is hardware that was specifically created to scale performance to a point where large-scale simulations can take advantage of these ever more accurate representations of the real world and so you can instrument the twin effectively to gain insight on its reality-based counterpart. Depending on the need, OVX servers can be configured into a pod of eight to 16 servers or into a super pod made up of a potentially unlimited number of pods. 

Twins in the metaverse

As we enter the metaverse age, it is critical we understand that when developing the metaverse for simulation, we create digital twins that remain digital twins. If we don’t, the simulations will fail and the result, be it autonomous robots, cars, or aircraft, will be problems that are not identified in simulation and are instead identified after a potentially catastrophic failure. 

Assuring digital twins remain accurate requires a very different kind of server, such as OVX, which NVIDIA announced at their GTC event and can be combined into pods or super pods. The result, as this technology advances, is a reality-adjacent digital twin metaverse and a far more capable resulting simulation environment. 

While we clearly are not there yet, the OVX server is a major stepping stone toward our simulated future. 

See more: AWS Introducing IoT TwinMaker

Subscribe to Data Insider

Learn the latest news and best practices about data science, big data analytics, artificial intelligence, data security, and more.

Similar articles

Get the Free Newsletter!

Subscribe to Data Insider for top news, trends & analysis

Latest Articles